Deep Learning Segmentation of Triple-Negative Breast Cancer (TNBC) Patient Derived Tumor Xenograft (PDX) and Sensitivity of Radiomic Pipeline to Tumor Probability Boundary

Abstract

Preclinical magnetic resonance imaging (MRI) is a critical component in a co-clinical research pipeline. Importantly, segmentation of tumors in MRI is a necessary step in tumor phenotyping and assessment of response to therapy. However, manual segmentation is time-intensive and suffers from inter- and intra- observer variability and lack of reproducibility. This study aimed to develop an automated pipeline for accurate localization and delineation of TNBC PDX tumors from preclinical T1w and T2w MR images using a deep learning (DL) algorithm and to assess the sensitivity of radiomic features to tumor boundaries. We tested five network architectures including U-Net, dense U-Net, Res-Net, recurrent residual UNet (R2UNet), and dense R2U-Net (D-R2UNet), which were compared against manual delineation by experts. To mitigate bias among multiple experts, the simultaneous truth and performance level estimation (STAPLE) algorithm was applied to create consensus maps. Performance metrics (F1-Score, recall, precision, and AUC) were used to assess the performance of the networks. Multi-contrast D-R2UNet performed best with F1-score = 0.948; however, all networks scored within 1–3% of each other. Radiomic features extracted from D-R2UNet were highly corelated to STAPLE-derived features with 67.13% of T1w and 53.15% of T2w exhibiting correlation ρ ≥ 0.9 (p ≤ 0.05). D-R2UNet-extracted features exhibited better reproducibility relative to STAPLE with 86.71% of T1w and 69.93% of T2w features found to be highly reproducible (CCC ≥ 0.9, p ≤ 0.05). Finally, 39.16% T1w and 13.9% T2w features were identified as insensitive to tumor boundary perturbations (Spearman correlation (−0.4 ≤ ρ ≤ 0.4). We developed a highly reproducible DL algorithm to circumvent manual segmentation of T1w and T2w MR images and identified sensitivity of radiomic features to tumor boundaries.

Effect and Prognosis Factors of Combining Laparoscopic Radical Resection of Colon Adenocarcinoma with Docetaxel Therapy in Treating Middle and Advanced Colon Adenocarcinoma

“…The latest report showed [ 11 ] that COAD incidence and mortality respectively rank 3 rd and 4 th in all malignant tumors, of which the incidence rose from 5 th to 3 rd in three years (2015 to 2018) according to the relevant study data, demonstrating that the incidence of COAD in China is increasing on a yearly basis. Surgical resection is still the main treatment of COAD [ 12 ], and the short-term efficacy of laparoscopic radical resection for COAD has been recognized. Clinical studies found that [ 13 , 14 ] performing laparoscopic radical resection to COAD patients who present with surgical indications can achieve the same results as an open surgery.…”

Deep Learning and Transfer Learning for Malaria Detection

“…Unfreeze allows us to choose which layers of your model to train at any given time by removing them from the freeze state. This is due to the fact that the early layers of our model will already be well trained in recognizing basic lines, patterns, and gradients, whereas the later layers (which will be more specific to our aim, such as identifying parasitemia) will necessitate further training [ 36 – 39 ]. By fine-tuning pretrained networks, we may utilize them to recognize classes that they were not trained on in the first place.…”

White Blood Cells Image Classification Based on Radiomics and Deep Learning

“…After image processing, transformation-based texture features reflect the texture and grayscale distribution within regions of interest in several locations [41]. Based on the segmented cell nucleus convex packet region, several features commonly used in these four types of texture features are extracted in this research [42][43] [44], which are: a) firstorder texture features: Variance, Kurtosis, Skewness; b) second-order texture features: GLCM [45]; c) higher-order texture features: Gray-Level Run-Length Matrix (GLRLM [46]), Gray-Level Size Zone Matrix (GLSZM [47]), Neighborhood Gray Tone Difference Matrix (NGTDM [48]); d) transform-based texture features: Median Robust Extended Local Binary Pattern (MRELBP [49]).…”